Symmetry

19 ms·

301.
by Symmetry 2y ago
Recoding America , a book about improving the quality of software used by the US federal government, had good things to say about how the UK did things and what we could learn from them.
302.
by Symmetry 2y ago
As CPU cores get larger and larger it makes sense to always keep looking for opportunities to decouple things. AMD went with separate schedulers in the Athalon three architectural overhauls ago and hasn't reversed their decision.
303.
by Symmetry 2y ago
Removing a core's SMT aka "hyperthreading" has some modest hardware savings but but biggest cost is that it makes testing and validation much more complicated. Given the existence of Intel's p-cores I'm not surpris
304.
by Symmetry 2y ago
The Shannon Limit is fundamental so you can't get around it directly. But yes, there are things you can do to receive information from further away. 1) A bigger dish is the most obvious one. 2) Use a lower bitrate to send more energy
305.
by Symmetry 2y ago
That reminds me of giving the advice, "Just do what comes naturally," to beginners. That's the last advice someone who has spent years building up an intuitive model needs, but it's terrible for beginners. https:/
306.
by Symmetry 2y ago
I have a sticker on my laptop of a yin-yang with DRY and YAGNI instead of the dots.
307.
by Symmetry 2y ago
Most Americans are only familiar with the US system and the completely socialized systems in Canada and Britain. Something like the French system isn't really politically on the table here.
308.
by Symmetry 2y ago
If you're talking about the study that everyone was talking about last week, that definition of "spare bedroom" includes a family of four with three bedrooms who gives each child their own bedroom when they could just have th
309.
by Symmetry 2y ago
If houses were just a commodity to be bought and sold we'd have much less of a problem. There were very few homeless a hundred years ago. Efforts to ban residential hotels aka "flophouses" were mostly an effort to remove
310.
by Symmetry 2y ago
Anthropic was founded by ex OpenAI employees who were concerned with the way it was being run and their language models are comparable, better for some things but worse than others. I also canceled my ChatGPT subscription and I will say I&
311.
by Symmetry 2y ago
This is not normal for being a corporate employee. This was certainly going to come out eventually and cause big problems, but to the extent Sam thinks AGI is around the corner he might not be playing the long game.
312.
by Symmetry 2y ago
Importantly if he just said publicly that that he wouldn't enforce the non-disparagement agreements that could be legally binding[1]. But if he just says he'll release people who he's, legally speaking, free to just not do t
313.
by Symmetry 2y ago
I'm sure there have been more expensive UX failures (Three Mile Island?) but if there's ever a Wikipedia list of them I'm sure this will be on it.
314.
by Symmetry 2y ago
SBIRs can be useful, but they can also be a trap where a company just keeps going after SBIRs and never takes the technology to market.
315.
by Symmetry 2y ago
Zvi has a pretty good summary of what happened: https://thezvi.wordpress.com/2024/05/22/do-not-mess-with-sca...
316.
by Symmetry 2y ago
I think there's only one part of that essay I disagree with: That SpaceX knows "How much propellant a Starship can carry to low Earth orbit". They're iterating on Starship. Falcon 9 started out with an LEO payload of 1
317.
by Symmetry 2y ago
It's important to remember that Apollo was one of Kennedy's signature political projects at the time he was assassinated, which was an important factor in its political viability.
318.
by Symmetry 2y ago
The name "RISC" is something we gave to a particular design philosophy, but not everything in that philosophy is summed up in the name. Classic RISC designs have 32 registers to make up for the lack of instructions that operate d
319.
by Symmetry 2y ago
The end result seems to be that the two entries from that old post that have seen the most success, x86 and ARM, are essentially the RISCiest of the CISCs and the CISCiest of the RISCs. A happy medium.
320.
by Symmetry 2y ago
Nobody has challenged it in court.
321.
by Symmetry 2y ago
Also a few IBM designs. An interesting thing is that IBM claimed that eDRAM was faster than SRAM for their particular design because at the scale they were using it the reduced size of the array increased speed more than eDRAM's acces
322.
by Symmetry 2y ago
That isn't true at all. The limited speed at which a signal can propagate itself across a chip and the added levels of muxing necessarily mean that there's a limit to how low the latency of a cache can be that's roughly prop
323.
by Symmetry 2y ago
Are you including address translation in the area you're ascribing to the L1$ there? I haven't looked at detailed area numbers for recent Intel designs but having equal area for L1$ and L2$ seems really weird to me based on numbe
324.
by Symmetry 2y ago
It is an alarmist take but that doesn't meant that people haven't believed it.
325.
by Symmetry 2y ago
Looking up an estimate of the brain's input bandwidth at 10 million bits per second and multiplying by the second or two a subliminal stimuli can continue to affect a person's behavior. This is a very crude estimate and probably
326.
by Symmetry 2y ago
I donate platelets with the Red Cross every month and I've got an app where they give me the blood pressure and hemoglobin level, and which used to tell me if I had Covid-19 antibodies before everybody did via vaccine or infection. Bef
327.
by Symmetry 2y ago
With each new IPCC report the chances of us keeping global warming below 1.5 C become more and more remote as emissions are locked in. But at the same time the really bad scenarios where we go well over 4 C also become more remote. Every
328.
by Symmetry 2y ago
The scenario I have in my head is that they had to override the safety team's objections to ship their new models before Google IO happened.
329.
by Symmetry 2y ago
A working memory that can preserve information indefinitely outside a particular context window and which can engage in multi-step reasoning that doesn't show up in its outputs. GPT4o's context window is 128k tokens which is somew
330.
by Symmetry 2y ago
Sounds great for small island nations.

More ›